Table 3 Accuracy (%) comparison on DREAM benchmarks.

From: English-focused CL-HAMC with contrastive learning and hierarchical attention for multiple-choice reading comprehension

Model

DREAM

BERT - large8

66.8

BERT - large + WAE16

69.0

XLNet - large35

72.0

ALBERT - xxlarge10

88.5

RoBERTa - large + MMM12

88.9

ALBERT - xxlarge + DUMA17

90.5

UDG - LLM27

92.07

ALBERT - xxlarge + CL-HAMC

90.8

  1. All values represent classification accuracy (%). Citations correspond to reference numbers in the main text. The proposed CL-HAMC model is highlighted in bold.
Back to article page